[IMPALA-3644] PlannerTest.java fails when run with Java 8

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Duplicate
• Affects Version/s: Impala 2.6.0
• Component/s: Frontend
• Assignee: Alexander Behm
• Reporter: Lars Volker
• Created: 31/May/16 12:13
• Updated: 19/Jun/18 17:32
• Resolved: 19/Jun/18 17:32

Description

twmarshall - I picked you a this looks like a predicate ordering issue, thinking you might have an idea what’s going on here; feel free to find another person or assign back to me if you're swamped.

PlannerTest.java:178 - testTpchNested fails on my local machine with a mismatch in the returned plan. The lines in error are the following:

Returned:

01:SCAN HDFS [tpch_nested_parquet.customer c]
   partitions=1/1 files=4 size=577.87MB
   predicates: !empty(c.c_orders)
*  predicates on o: !empty(o.o_lineitems), o_orderstatus = 'F'
   predicates on l1: l1.l_receiptdate > l1.l_commitdate
   predicates on l3: l3.l_receiptdate > l3.l_commitdate

Expected:

01:SCAN HDFS [tpch_nested_parquet.customer c]
   partitions=1/1 files=4 size=577.87MB
   predicates: !empty(c.c_orders)
*  predicates on o: o_orderstatus = 'F', !empty(o.o_lineitems)
   predicates on l1: l1.l_receiptdate > l1.l_commitdate
   predicates on l3: l3.l_receiptdate > l3.l_commitdate

It looks like the predicate order is reversed. I also noticed IMPALA-3643 on the same machine, which looks to me like it could be related to Java HashMap iterating over elements in an unexpected order on my machine/JVM. Could this also be the case here? I couldn't figure out where the order of the predicates is determined. There are multiple tests failing in PlannerTest.java for the same reason.

My HEAD is at * 7167950 - Remove redundant test in test_avro_schema_resolution.py (3 days ago) <Alex Behm>.

Issue Links

• duplicates: IMPALA-5344 Frontend tests do not work with Java 8